Low-altitude observation method for concentration of suspended sediment in seawater

Abstract

The invention relates to the technical field of marine hydrology monitoring, and discloses a low-altitude observation method for seawater suspended sediment concentration, which comprises the following steps of: carrying digital photographic equipment by using a low-altitude observation platform, acquiring a seawater suspended sediment concentration image, simultaneously carrying out water sample collection on a photographic sea area, positioning by using a global satellite positioning system, and measuring the suspended sediment content by using a filtering method. The obtained seawater suspended sediment concentration image is preprocessed, and automatic splicing of the image and automatic imaging of the orthoimage are achieved. And based on an optimized artificial neural network algorithm, training by using the processed image data and actually measured suspended sediment concentration data, establishing a seawater suspended sediment concentration inversion model, and evaluating the model precision. According to the method, a high-resolution image is obtained by using a low-altitude observation platform, suspended sediment concentration inversion is carried out by optimizing an artificial neural network model, the method has the advantages of high precision, low cost, flexibility, timeliness and the like, and the defects of satellite remote sensing can be effectively overcome.

Description

Technical Field

[0001] The invention belongs to the technical field of ocean hydrological monitoring, and in particular relates to a low-altitude observation method for suspended sediment concentration in seawater. Background Art

[0002] Suspended sediment is one of the important elements of marine hydrology and has an extremely important impact on human activities. Suspended sediment has a strong adsorption capacity. It is not only a carrier of nutrients and organic matter, but also adsorbs heavy metals and toxic wastes, and plays an important role in the migration and circulation of pollutants. Suspended sediment significantly reduces the light transmittance and photosynthesis efficiency of water bodies through effects such as refraction and scattering of light, affecting the primary productivity of the ocean and the types, quantity and distribution of phytoplankton and benthic organisms, and has different degrees of influence on the growth of algae, the absorption of nutrients and population succession. Suspended sediment controls the transport process of nutrients to a certain extent, resulting in the uneven distribution of nutrients in the water body in time and space, which is one of the reasons why red tides show seasonal and regional characteristics. Suspended sediment is closely related to changes in coastal topography, siltation of port channels, and changes in scouring and silting around buildings. It often involves port site planning and layout, channel dredging, selection of mud dumping areas, coastline erosion, safety of marine structure foundations, breakwaters, land reclamation, and bridge construction.

[0003] Suspended sediment plays an important role in the migration and circulation of pollutants, marine biochemical cycles, carbon cycles, light transmittance of seawater, coastal erosion and siltation, port channels, coastal engineering, etc., and is therefore the focus of multidisciplinary attention. A large number of observations and studies on suspended sediment concentration have been conducted at home and abroad based on shore-based, ship-based, submarine-based, and seabed-based methods using traditional water sampling and filtration methods, optical conversion methods, and acoustic inversion methods. These methods are usually based on point or line observations, which are difficult to meet the needs of rapid acquisition of suspended sediment concentration in large areas of seawater. The development of satellite remote sensing technology has made the space-based seawater suspended sediment concentration inversion technology more mature, providing a fast and large-area remote sensing method for seawater suspended sediment concentration observations. Previous researchers have conducted a large number of seawater suspended sediment concentration inversion observations based on satellite remote sensing technology, making the seawater suspended sediment concentration satellite remote sensing inversion method quite mature. However, satellites have a certain orbital period, and the observation time is limited by the re-entry period. At the same time, the satellite is high above the sea surface, the signal propagation distance is long, and it is greatly affected by the curvature of the earth and the weather. The image spatial resolution is low, making it difficult to accurately monitor small-scale sea areas. Although remote sensing using drones equipped with multispectral or hyperspectral cameras makes up for some of the shortcomings of satellite remote sensing, the high prices of multispectral or hyperspectral cameras limit their large-scale application.

[0004] In view of this, we proposed a method of using a low-altitude observation platform to obtain low-altitude images of seawater suspended sediment concentration. Based on an artificial neural network, we trained the seawater suspended sediment concentration image data and the measured seawater suspended sediment concentration, established a seawater suspended sediment concentration inversion model, and realized a low-altitude observation method of seawater suspended sediment concentration to solve the above problems. Summary of the invention

[0005] The present invention aims to solve the limitation of the above-mentioned prior art suspended sediment concentration observation method in the aspect of rapid acquisition of large-scale small-scale sea areas. Although satellite remote sensing provides a remote sensing means, it is limited by the revisit cycle and image resolution, and it is difficult to accurately monitor small-scale sea areas. Although drone remote sensing technology can make up for some of the shortcomings of satellite remote sensing, the high equipment cost limits its wide application.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A low-altitude observation method for suspended sediment concentration in seawater comprises the following steps:

[0008] S1. With the help of a low-altitude observation platform, digital photography is used to obtain images of suspended sediment concentration in the target sea area. The images obtained can be stored on a chip memory card to be downloaded or transferred to a computer workstation for subsequent processing; water sampling and global satellite positioning are carried out simultaneously in the photographed sea area. The collected water samples are used to determine the suspended sediment content by filtration method, and the collected water samples are divided into modeling samples and verification samples;

[0009] S2. Preprocess the acquired seawater suspended sediment concentration images, including mean filtering, geo-registration, and angle correction, and generate image data for further analysis through automatic stitching and orthophoto processing;

[0010] S3. According to the collected suspended sediment concentration image data and the measured suspended sediment concentration data, an artificial neural network is constructed. Based on the optimization of the artificial neural network algorithm, the artificial neural network is trained using the seawater suspended sediment concentration image data and the measured suspended sediment concentration data, the network parameters are adjusted, the model performance is optimized, and a low-altitude inversion model for seawater suspended sediment concentration is established;

[0011] S4. Use the remote sensing suspended sediment concentration inversion model and measured verification samples to verify the low-altitude inversion model of seawater suspended sediment concentration, evaluate the accuracy of the seawater suspended sediment concentration inversion model, and adjust and optimize the model based on the evaluation results;

[0012] S5. Apply the low-altitude inversion model of seawater suspended sediment concentration to the target sea area image data to invert the low-altitude observation results of seawater suspended sediment concentration.

[0013] Preferably, in step S1, the low-altitude observation platform is a platform with a certain height capable of low-altitude photography and obtaining digital images of suspended sediment concentration in seawater, including but not limited to drones, airships, coastal heights, structures, offshore platforms, and ships.

[0014] Preferably, in step S1, the digital photography includes a high-resolution visible light digital camera such as a smart phone or a camera.

[0015] Preferably, in step S1, when obtaining the image of suspended sediment concentration in seawater by digital photography, a fixed angle is used or the angle is corrected to an orthographic angle.

[0016] Preferably, in step S2, the specific implementation method of geo-registration during the seawater suspended sediment concentration image processing is:

[0017] Geo-reference: The camera system is equipped with a global satellite positioning system, and real-time dynamic carrier phase difference is used for spatial positioning, or characteristic objects are selected and spatially positioned using the global satellite positioning system to obtain the longitude and latitude of the objects. The coordinate points of the characteristic objects and the pixel points on the image are combined to perform geo-reference using ArcGIS software. If there are no characteristic objects in the observation area, they are manually marked.

[0018] The specific implementation method of angle correction in seawater suspended sediment concentration image processing is:

[0019] Monochromatic liquid was used for testing under the same light source. The camera was used to shoot at different angles from 20° to 90°. Matlab software was used to extract the RED, GREEN and BLUE band values ​​of the camera image. Function fitting was performed between the different camera inclination angles and the camera band values ​​at the same position. The angle correction equation was obtained as follows: C X =13*λ*C i / 1000+λ, where C X is the corrected band value, C i is the uncorrected band value, λ is the angle between the camera and the horizontal plane;

[0020] The angle correction equation is used to correct the RED, GREEN and BLUE band values ​​to the positive angle. The corrected RED, GREEN and BLUE band values ​​are: C R =13*λ*C r / 1000+λ;

[0021] C G =13*λ*C g / 1000+λ;C b =13*λ*C b / 1000+λ.

[0022] Preferably, based on the artificial neural network algorithm, the artificial neural network training is carried out using the seawater suspended sediment concentration image band values ​​and the measured suspended sediment concentration to establish a low-altitude inversion model of seawater suspended sediment concentration; the artificial neural network is a BP neural network model containing an input layer, two hidden layers and an output layer.

[0023] Preferably, Matlab is used to construct a neural network model including an input layer, a hidden layer, and an output layer. The input layer is RGB band data extracted from camera images, and the output layer is suspended sediment concentration data. The band data of each pixel is input, and the suspended sediment concentration is the corresponding output. The neural network algorithm training data includes two parts, namely input data and output data. The two parts of data are combined into data pairs, and 70% of the measured seawater suspended sediment concentration data and the processed and corrected image band RGB data pairs are used as training data sets, and the remaining data pairs are used as verification test data sets.

[0024] Use mapminmax function to normalize input data and output data to the interval [0,1]; use feedforwardnet function to build a neural network; the hidden layer uses tansig activation function, and the number of neurons in the hidden layer is set through experimental comparison and analysis. The number of neurons in the first hidden layer is 12, and the number of neurons in the second hidden layer is 7; the output layer uses linear activation function purelin;

[0025] The training algorithm selects trainlm to train the neural network and sets the training parameters, including the maximum number of training times 1000, the learning rate 0.01, and the training target error 10. -6 ,After optimization, 5-fold cross validation and L2 regularization are used to avoid overfitting and optimize the neural network parameters.

[0026] Preferably, in step S4, the low-altitude inversion model of seawater suspended sediment concentration is verified by comparing it with the measured suspended sediment concentration and the suspended sediment concentration inverted by remote sensing, and the accuracy of the suspended sediment concentration inversion model is evaluated by three indicators: mean square error (MSE), root mean square error (RMSE), and mean absolute error (MAE). The formula is as follows:

[0027]

[0028]

[0029]

[0030] Among them, y i is the measured value, is the model prediction value, and n is the number of samples.

[0031] Preferably, the low-altitude inversion model of seawater suspended sediment concentration is suitable for seawater, lakes and reservoirs.

[0032] Compared with the prior art, the technical effects and advantages of the present invention are:

[0033] This low-altitude observation method of seawater suspended sediment concentration takes advantage of the low-altitude observation platform, gets closer to the seawater to obtain information on seawater suspended sediment concentration, and realizes low-altitude observation of seawater suspended sediment concentration. It has the advantages of high resolution, high accuracy, low cost, fast and efficient, flexible, low risk, and repeatability, and is suitable for small-scale sea area observations.

[0034] The low-altitude observation method of seawater suspended sediment concentration uses an optimized neural network algorithm. By setting the number of hidden layer neurons, the optimized neural network algorithm constructs more suitable inversion parameters for seawater suspended sediment concentration in complex water bodies, thereby improving the suitability and reliability of the model.

[0035] The low-altitude observation method of suspended sediment concentration in seawater uses the bands and water body characteristics in the suspended sediment concentration image to invert the suspended sediment concentration, making up for the deficiency of using spectral characteristics to invert the suspended sediment concentration in seawater. The invention can accurately describe, predict and evaluate the changes in suspended sediment movement, which is conducive to promoting the ecological environment protection of nearshore waters and scientific development of coastal areas, and has broad application prospects in the field of marine hydrological observation. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a flow chart of the low-altitude observation method of seawater suspended sediment concentration of the present invention;

[0037] Figure 2 This is a verification diagram of the inversion result of the seawater suspended sediment concentration of the present invention;

[0038] Figure 3 This is a diagram of the inversion results of different seawater suspended sediment concentrations according to the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] The following is combined with Figure 1-3 To further explain this application in detail,

[0041] The present application embodiment discloses a method for low-altitude observation of suspended sediment concentration in seawater, comprising the following steps:

[0042] S1. Data collection: Select a suitable low-altitude observation platform and prepare for observation. Low-altitude observation platforms include drones, airships, coastal highlands, structures, offshore platforms or ships, etc., which have a certain height and can be used for low-altitude photography and acquisition of digital images of suspended sediment concentration in seawater;

[0043] Use digital photography to obtain images of suspended sediment concentration in seawater. The images obtained can be stored on a chip memory card and can be downloaded or transferred to a computer workstation for subsequent processing. Digital photography includes high-resolution visible light digital cameras such as smart phones and cameras;

[0044] When using digital photography to obtain images of suspended sediment concentration in seawater, a fixed angle or an angle correction to an orthographic angle is used, and water samples are collected and global satellite positioning is performed in the photographed sea area. The collected water samples are filtered to determine the suspended sediment content. The collected water samples are divided into training samples and verification samples.

[0045] S2. Image processing: Process the acquired seawater suspended sediment concentration images. Since the acquired seawater suspended sediment concentration images are affected by factors such as shooting angle, light intensity, and seawater fluctuations, the collected image data needs to be corrected, including mean filtering, georeferencing, and angle correction to reduce noise and errors and improve data quality, so as to more accurately reflect the information on seawater suspended sediment concentration. Through automatic stitching and orthophoto processing, image data that can be used for further analysis is generated. The specific implementation method is:

[0046] (1) Mean filtering. The resolution of suspended sediment concentration images obtained at low altitude is much higher than that of remote sensing, and even small changes in the sea surface can be identified. Since the fluctuation of sea water will cause the surface of the sea to fluctuate, when shooting at a certain angle, the image will appear rippled. It is necessary to perform mean filtering on the image and interpolate the image null values ​​to reduce the error.

[0047] (2) Geo-reference. Select characteristic features, use the global satellite positioning system for spatial positioning, and obtain the longitude and latitude of the features. Combine the coordinate points of the characteristic features with the pixel points on the image and use ArcGIS software for geo-reference. If there are no characteristic features in the observation area, they can be marked manually.

[0048] (3) Angle correction. When collecting suspended sediment concentration image data, different shooting angles will lead to differences in the band values ​​of the image, which need to be corrected. Monochromatic liquid is used for testing under the same light source. The camera is used to shoot at different angles from 20° to 90°. The RED, GREEN and BLUE band values ​​of the camera image are extracted using Matlab software. The function of different camera inclination angles and the camera band values ​​at the same position is fitted to obtain the angle correction equation: C X =13*λ*C i / 1000+λ, where C X is the corrected band value, C i is the uncorrected band value, λ is the angle between the camera and the horizontal plane; the RED, GREEN and BLUE band values ​​are corrected to the positive angle using the angle correction equation. The corrected RED, GREEN and BLUE band values ​​are: C R =13*λ*C r / 1000+λ;C G =13*λ*C g / 1000+λ;C b =13*λ*C b / 1000+λ.

[0049] S3. Model establishment and training: Based on the collected suspended sediment concentration image data and measured suspended sediment concentration data, an artificial neural network with input layer, hidden layer and output layer is constructed. The suspended sediment concentration image data and the corresponding measured data are used to train the neural network, adjust the network parameters, and optimize the model performance.

[0050] Based on the artificial neural network algorithm, the artificial neural network training is carried out using the seawater suspended sediment concentration image band values ​​and the measured suspended sediment concentration data to establish a low-altitude inversion model for seawater suspended sediment concentration. Matlab is used to construct a neural network model including an input layer, a hidden layer, and an output layer. The input layer of this embodiment is the RGB band data extracted from the camera image, and the output layer is the suspended sediment concentration data. Specifically, the band data of each pixel point is the input, and the suspended sediment concentration is the corresponding output. The neural network algorithm training data includes two parts, namely input data and output data. The two parts of the data are composed of data pairs. 70% of the measured seawater suspended sediment concentration data and the processed and corrected image band RGB data pairs are used as training data sets, and the remaining data pairs are used as verification test data sets. The mapminmax function is used to normalize the input data and output data to the [0,1] interval. The feedforwardnet function is used to construct a neural network. The hidden layer uses the tansig activation function, and the number of neurons in the hidden layer is set through experimental comparative analysis. The number of neurons in the first hidden layer is 12, and the number of neurons in the second hidden layer is 7. The output layer uses the linear activation function purelin.

[0051] Select trainlm as the training algorithm to train the neural network. Set the training parameters, including the maximum number of training times (1000), learning rate (0.01), training target error (10 -6 ), using 5-fold cross validation and L2 regularization to avoid overfitting and optimize neural network parameters. After training, the training results of each group tend to be stable and meet the error requirements. The accuracy of the model inverting suspended sediment concentration in complex water bodies is improved.

[0052] S4. Model verification and accuracy assessment: The remote sensing suspended sediment concentration inversion model and the measured verification samples were used to verify the low-altitude inversion model of seawater suspended sediment concentration. The mean square error (MSE), root mean square error (RMSE) and mean absolute error (MAE) were used to evaluate the accuracy of the seawater suspended sediment concentration inversion model.

[0053]

[0054]

[0055]

[0056] Among them, y i is the measured value, is the model prediction value, and n is the number of samples.

[0057] In this embodiment, the mean square error (MSE) is 120.04, the root mean square error (RMSE) is 10.95, the mean absolute error (MAE) is 7.04, and the model inversion accuracy is 89.75%. Figure 2 In order to compare the low-altitude inversion results of seawater suspended sediment concentration with the inversion accuracy of satellite remote sensing data, the low-altitude observation accuracy has been significantly improved;

[0058] S5. Application of results: The low-altitude inversion model of seawater suspended sediment concentration is applied to the image data of the target sea area, and the low-altitude observation results of seawater suspended sediment concentration are inverted.

[0059] The inversion model is used to observe the suspended sediment concentration in seawater. Figure 3 The inversion results of suspended sediment concentration images of different concentrations of seawater in Haizhou Bay were taken with a mobile phone. a1, a2, and a3 are the original camera images, which are seawater images with different suspended sediment concentrations. a1 is a low-concentration water body, a2 is a high-concentration water body, and a3 is a medium-high concentration water body. b1, b2, and b3 are the corresponding inversion results of suspended sediment concentration in seawater. It can be seen from the inverted image that the suspended sediment concentration ranges from 50 to 75 mg / l, and the distribution of suspended sediment is consistent with the measured results.

[0060] The present invention discloses a low-altitude observation method for suspended sediment concentration in seawater. The method uses a low-altitude observation platform, such as an unmanned aerial vehicle, an airship, a coastal highland, a structure, an offshore platform or a ship, to obtain high-resolution image data of suspended sediment concentration in seawater closer to the seawater, thereby improving the observation accuracy of suspended sediment concentration. Compared with satellite remote sensing and multi-spectral cameras and hyperspectral cameras, the digital photography equipment (such as smart phones and cameras) used in this method is low-cost and convenient for large-scale promotion and application. The method can quickly obtain suspended sediment concentration information in a large area of ​​sea, thereby improving the observation efficiency. With the help of a low-altitude observation platform, the observed sea area and observation time can be adjusted at any time as needed, and the method has strong flexibility. Compared with traditional observation methods, the low-altitude observation platform has lower risks and is suitable for complex sea conditions. Low-altitude observations can be repeated quickly, which is convenient for dynamic monitoring of changes in suspended sediment concentration.

[0061] This low-altitude observation method of seawater suspended sediment concentration effectively makes up for the shortcomings of existing technologies in inverting seawater suspended sediment concentration. The model is more accurate, which improves the timeliness, accuracy and flexibility of seawater suspended sediment concentration observation and realizes efficient observation of seawater suspended sediment concentration in small-scale sea areas.

[0062] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A low-altitude observation method for suspended sediment concentration in seawater, characterized in that: The steps include: S1. With the help of a low-altitude observation platform, digital photography is used to obtain images of suspended sediment concentration in the target sea area. The images obtained can be stored on a chip memory card to be downloaded or transferred to a computer workstation for subsequent processing; water sampling and global satellite positioning are carried out simultaneously in the photographed sea area. The collected water samples are used to determine the suspended sediment content by filtration method, and the collected water samples are divided into modeling samples and verification samples; S2. Preprocess the acquired seawater suspended sediment concentration images, including mean filtering, geo-registration, and angle correction, and generate image data for further analysis through automatic stitching and orthophoto processing; S3. According to the collected suspended sediment concentration image data and the measured suspended sediment concentration data, an artificial neural network is constructed. Based on the optimization of the artificial neural network algorithm, the artificial neural network is trained using the seawater suspended sediment concentration image data and the measured suspended sediment concentration data, the network parameters are adjusted, the model performance is optimized, and a low-altitude inversion model for seawater suspended sediment concentration is established; S4. Use the remote sensing suspended sediment concentration inversion model and measured verification samples to verify the low-altitude inversion model of seawater suspended sediment concentration, evaluate the accuracy of the seawater suspended sediment concentration inversion model, and adjust and optimize the model based on the evaluation results; S5. Apply the low-altitude inversion model of seawater suspended sediment concentration to the target sea area image data to invert the low-altitude observation results of seawater suspended sediment concentration.

2. A method for low-altitude observation of suspended sediment concentration in seawater according to claim 1, characterized in that: In step S1, the low-altitude observation platform is a platform with a certain height capable of low-altitude photography and obtaining digital images of suspended sediment concentration in seawater, including but not limited to drones, airships, coastal highlands, structures, offshore platforms or ships.

3. The low-altitude observation method for suspended sediment concentration in seawater according to claim 1, characterized in that: In step S1, digital photography includes high-resolution visible light digital cameras such as smart phones and cameras.

4. The low-altitude observation method for suspended sediment concentration in seawater according to claim 1, characterized in that: In step S1, when the image of suspended sediment concentration in seawater is obtained by digital photography, a fixed angle is used or the angle is corrected to an orthographic angle.

5. The low-altitude observation method for suspended sediment concentration in seawater according to claim 1, characterized in that: In step S2, the specific implementation method of geo-registration during the processing of seawater suspended sediment concentration image is as follows: Geo-reference: The camera system is equipped with a global satellite positioning system, and real-time dynamic carrier phase difference is used for spatial positioning, or characteristic objects are selected and spatially positioned using the global satellite positioning system to obtain the longitude and latitude of the objects. The coordinate points of the characteristic objects and the pixel points on the image are combined to perform geo-reference using ArcGIS software. If there are no characteristic objects in the observation area, they are manually marked. The specific implementation method of angle correction in seawater suspended sediment concentration image processing is: Monochromatic liquid was used for testing under the same light source. The camera was used to shoot at different angles from 20° to 90°. Matlab software was used to extract the RED, GREEN and BLUE band values ​​of the camera image. Function fitting was performed between the different camera inclination angles and the camera band values ​​at the same position. The angle correction equation was obtained as follows: C X =13*λ*C i / 1000+λ, where C X is the corrected band value, C i is the uncorrected band value, λ is the angle between the camera and the horizontal plane; The angle correction equation is used to correct the RED, GREEN and BLUE band values ​​to the positive angle. The corrected RED, GREEN and BLUE band values ​​are: C R =13*λ*C r / 1000+λ; C G =13*λ*C g / 1000+λ;C B =13*λ*C b / 1000+min.

6. The low-altitude observation method for suspended sediment concentration in seawater according to claim 1, characterized in that: Based on the artificial neural network algorithm, the artificial neural network training was carried out using the seawater suspended sediment concentration image band values ​​and the measured suspended sediment concentration to establish a low-altitude inversion model for seawater suspended sediment concentration; the artificial neural network was a BP neural network model containing an input layer, two hidden layers and an output layer.

7. A method for low-altitude observation of suspended sediment concentration in seawater according to claim 6, characterized in that: Matlab was used to construct a BP neural network model including input layer, hidden layer and output layer. The input layer is RGB band data extracted from camera images, and the output layer is suspended sediment concentration data. The band data of each pixel is input, and the suspended sediment concentration is the corresponding output. The neural network algorithm training data includes two parts, namely input data and output data. The two parts of data are combined into data pairs. 70% of the measured seawater suspended sediment concentration data and the processed and corrected image band RGB data pairs are used as training data sets, and the remaining data pairs are used as verification test data sets. Use mapminmax function to normalize input data and output data to the interval [0,1]; use feedforwardnet function to build a neural network; the hidden layer uses tansig activation function, and the number of neurons in the hidden layer is set through experimental comparison and analysis. The number of neurons in the first hidden layer is 12, and the number of neurons in the second hidden layer is 7; the output layer uses linear activation function purelin; The training algorithm selects trainlm to train the neural network and sets the training parameters, including the maximum number of training times 1000, the learning rate 0.01, and the training target error 10. -6 ,After optimization, 5-fold cross validation and L2 regularization are used to avoid overfitting and optimize the neural network parameters.

8. The low-altitude observation method for suspended sediment concentration in seawater according to claim 1, characterized in that: In step S4, the low-altitude inversion model of seawater suspended sediment concentration is verified by comparing it with the measured suspended sediment concentration and the suspended sediment concentration inverted by remote sensing, and the accuracy of the suspended sediment concentration inversion model is evaluated by three indicators: mean square error (MSE), root mean square error (RMSE), and mean absolute error (MAE). The formula is as follows: Among them, y i is the measured value, is the model prediction value, and n is the number of samples.

9. The low-altitude observation method for suspended sediment concentration in seawater according to claim 1, characterized in that: The low-altitude inversion model of seawater suspended sediment concentration is applicable to seawater, lakes and reservoirs.

Images